Re: Hermits and Cranks (fwd)

From: Eliezer S. Yudkowsky (
Date: Tue Feb 19 2002 - 09:03:02 MST

More from the nonexistent paper, "Special difficulties consequent to research
on generally intelligent supersystems":

I'm not sure you could legitimately call anyone working in AI a "crank". Take
Alan Grimes. I am willing to argue that he is an idiot and that his theory of
AI, insofar as it can be deciphered, is junk. I'm not sure that I could get
away with calling him an "AI crank" or an "AI crackpot", because right now the
state of the field is such that those researchers who think they have general
theories of intelligence are all arguing with each other, and those who don't
think so are bemoaning the fact that the field has still failed to yield any
general theory of intelligence. If someone's crackpot theory fails to be
readily shot down by established theory, due to lack of same, is it really a
crackpot theory? A stupid theory, yes, a wrong theory, yes, but crackpot?

Physics crackpots generally have theories that manage to contradict
established experiments and well-confirmed theories. In AI you may be able to
say that a theory is dumbfoundingly oversimplified and that it amounts to
recycled classical AI from the 50s and 60s, because this is what amateurs
automatically come up when they try to think about AI, but you don't have the
scientific base to call it crackpot. People with doctorates are also still
recycling 60s-era classical AI. I happen to think that Christopher
McKinstry's GAC is probably the worst AI theory in human history, and yet as
far as the almost-nonexistent body of established theory is concerned, a
sufficiently large database of yes-or-no questions might indeed cough up
general intelligence - who knows? The only person who I would really feel
confident in calling an AI crackpot is Mentifex.

-- -- -- -- --
Eliezer S. Yudkowsky
Research Fellow, Singularity Institute for Artificial Intelligence

This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 13:37:40 MST